Self-Supervised Learning Strategies for Jet Physics

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Rieck, Patrick, Cranmer, Kyle, Dreyer, Etienne, Gross, Eilam, Kakati, Nilotpal, Kobylanskii, Dmitrii, Merz, Garrett W., Soybelman, Nathalie
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912275041878016
author Rieck, Patrick
Cranmer, Kyle
Dreyer, Etienne
Gross, Eilam
Kakati, Nilotpal
Kobylanskii, Dmitrii
Merz, Garrett W.
Soybelman, Nathalie
author_facet Rieck, Patrick
Cranmer, Kyle
Dreyer, Etienne
Gross, Eilam
Kakati, Nilotpal
Kobylanskii, Dmitrii
Merz, Garrett W.
Soybelman, Nathalie
contents We extend the re-simulation-based self-supervised learning approach to learning representations of hadronic jets in colliders by exploiting the Markov property of the standard simulation chain. Instead of masking, cropping, or other forms of data augmentation, this approach simulates pairs of events where the initial portion of the simulation is shared, but the subsequent stages of the simulation evolve independently. When paired with a contrastive loss function, this naturally leads to representations that capture the physics in the initial stages of the simulation. In particular, we force the hard scattering and parton shower to be shared and let the hadronization and interaction with the detector evolve independently. We then evaluate the utility of these representations on downstream tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2503_11632
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Self-Supervised Learning Strategies for Jet Physics
Rieck, Patrick
Cranmer, Kyle
Dreyer, Etienne
Gross, Eilam
Kakati, Nilotpal
Kobylanskii, Dmitrii
Merz, Garrett W.
Soybelman, Nathalie
High Energy Physics - Phenomenology
High Energy Physics - Experiment
We extend the re-simulation-based self-supervised learning approach to learning representations of hadronic jets in colliders by exploiting the Markov property of the standard simulation chain. Instead of masking, cropping, or other forms of data augmentation, this approach simulates pairs of events where the initial portion of the simulation is shared, but the subsequent stages of the simulation evolve independently. When paired with a contrastive loss function, this naturally leads to representations that capture the physics in the initial stages of the simulation. In particular, we force the hard scattering and parton shower to be shared and let the hadronization and interaction with the detector evolve independently. We then evaluate the utility of these representations on downstream tasks.
title Self-Supervised Learning Strategies for Jet Physics
topic High Energy Physics - Phenomenology
High Energy Physics - Experiment
url https://arxiv.org/abs/2503.11632